[U-Boot] [PATCH V2 3/8] ehci-hcd.c, musb_core, usb.h: Add USB_DMA_MINALIGN define for cache alignment

[Marek Vasut <marex@denx.de>]

Dear Ilya Yanok,

> Dear Marek,
> 
> On Sun, Jul 15, 2012 at 6:59 PM, Marek Vasut <marex@denx.de> wrote:
> > > index 04300be..59039f4 100644
> > > --- a/drivers/usb/host/ehci-hcd.c
> > > +++ b/drivers/usb/host/ehci-hcd.c
> > > @@ -34,7 +34,10 @@ struct ehci_hccr *hccr;    /* R/O registers, not
> > > need
> > 
> > for
> > 
> > > volatile */ volatile struct ehci_hcor *hcor;
> > > 
> > >  static uint16_t portreset;
> > > 
> > > -static struct QH qh_list __attribute__((aligned(32)));
> > > +DEFINE_ALIGN_BUFFER(struct QH, qh_list, 1, USB_DMA_MINALIGN);
> > > +
> > > +#define ALIGN_END_ADDR(type, ptr, size)                      \
> > > +     ((uint32_t)(ptr) + roundup((size) * sizeof(type),
> > 
> > USB_DMA_MINALIGN))
> > 
> > >  static struct descriptor {
> > >  
> > >       struct usb_hub_descriptor hub;
> > > 
> > > @@ -207,8 +210,9 @@ static int
> > > 
> > >  ehci_submit_async(struct usb_device *dev, unsigned long pipe, void
> > > 
> > > *buffer, int length, struct devrequest *req)
> > > 
> > >  {
> > > 
> > > -     static struct QH qh __attribute__((aligned(32)));
> > > -     static struct qTD qtd[3] __attribute__((aligned (32)));
> > > +     /* for some reason this doesn't work with non-static qh */
> > > +     DEFINE_ALIGN_BUFFER(struct QH, qh, 1, USB_DMA_MINALIGN);
> > > 
> > > diff --git a/drivers/usb/host/ehci-hcd.c b/drivers/usb/host/ehci-hcd.c
> > 
> > For some reason this doesn't work, but for obvious reason this is NAK.
> > The rest
> > is OK, but this is not. Let's investigate.
> 
> But this was static since the initial EHCI support introduction... the
> patch just preserves the previous behavior.
> I agree we need to find the reason it doesn't work with on-stack qh and fix
> it but it's another problem, not the one these series try to address.

But this is like covering bugs ... the rest of the patches are OK, it's just 
this little detail that annoys me.

> > > +     ALLOC_ALIGN_BUFFER(struct qTD, qtd, 3, USB_DMA_MINALIGN);
> > > 
> > >       int qtd_counter = 0;
> > >       
> > >       volatile struct qTD *vtd;
> > 
> > [...]
> > 
> > I'll poke into this hopefully later today, but if you can try finding
> > something
> > until then, that'd be great :)
> 
> Sorry, I've already lost too much time with this and I won't return to this
> in the near future.

Dunno how much this might motivate you, but this is about my thought now:

http://farm8.staticflickr.com/7138/7456661744_331a4f3535.jpg

I'll try taking over from here though. Thanks for the set and for all your 
investment into this!

> Regards, Ilya.

Best regards,
Marek Vasut